Skip to content

Lower precision in effective token rendering functions#24216

Closed
Copilot wants to merge 2 commits intomainfrom
copilot/lower-precision-tokens-rendering
Closed

Lower precision in effective token rendering functions#24216
Copilot wants to merge 2 commits intomainfrom
copilot/lower-precision-tokens-rendering

Conversation

Copy link
Copy Markdown
Contributor

Copilot AI commented Apr 3, 2026 •
edited
Loading

Effective tokens are computed estimates (weighted token counts × model multipliers), not exact values. Displaying results with unnecessary precision implies false accuracy.

Changes

  • formatET (JS) — round values in [10, 1000) to nearest 10; values < 10 remain exact; K range unchanged; M range rounded to nearest whole M
  • formatTokens (Go) — same rounding applied throughout all ranges
Before → After
"42"    → "40"
"123"   → "120"
"900"   → "900"   (multiples of 10 unaffected)
"999"   → "1K"    (rounds up across threshold)
"1.2K"  → "1.2K"  (K range unchanged)
"1.2M"  → "1M"    (M range rounded to nearest whole M)
"7.8M"  → "8M"
"12.3M" → "12M"

@pelikhan
Copy link
Copy Markdown
Collaborator

pelikhan commented Apr 3, 2026

@copilot range is common around 1M - 10M

Copy link
Copy Markdown
Contributor Author

Copilot AI commented Apr 3, 2026

@copilot range is common around 1M - 10M

Good point. M-range values now round to the nearest whole million (e.g. "1.2M" → "1M", "7.8M" → "8M"), matching the precision level appropriate for the common 1M–10M range. Committed in 487bb47.

Copilot AI requested a review from pelikhan April 3, 2026 02:38
@pelikhan pelikhan closed this Apr 3, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@pelikhan pelikhan Awaiting requested review from pelikhan

Assignees

@pelikhan pelikhan

Copilot code review Copilot

Labels

None yet

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

2 participants

@pelikhan